Summary: Qualcomm introduced Snapdragon X2 Plus at CES on January 5, 2026 as 6-core and 10-core 3nm PC chips carrying the flagship’s 80-TOPS NPU into the $799-to-$1,299 mainstream band. The tier could normalize on-device AI on Windows Arm laptops, but early benchmarks used pre-production hardware, Apple’s M5 estimate is higher, and emulation and anti-cheat compatibility still have gaps.
January 5, 2026, CES media day: at its Snapdragon X Series event in the Wynn, Las Vegas, Qualcomm introduced the Snapdragon X2 Plus — the mainstream rung that completes the family the X2 Elite and X2 Elite Extreme opened at Snapdragon Summit in September 2025. It arrives in 6-core and 10-core trims on a 3nm process, and strapped to it is the very same 80-TOPS Hexagon NPU the flagship carries. HP, Lenovo, and ASUS announced first devices at CES; CNET reports some models reaching shelves by the end of March, the rest landing across the first half of 2026.
For developers and product teams, the chip’s point is not the spec sheet. It is that a genuinely capable NPU is about to move into the $799-to-$1,299 band where most PCs actually sell — the very segment Windows on Arm has spent two years failing to crack.
What Ships: Two Core Counts, One 80-TOPS NPU
The X2 Plus ships as the 6-core X2P-42-100 and the 10-core X2P-64-100. Both ride the second-generation Oryon CPU architecture on 3nm (down from the 4nm of the previous Snapdragon X Plus), peak around 4.0 GHz, and carry 22MB and 34MB of cache respectively. Graphics come from an Adreno X2-45 GPU — 0.90 GHz on the 6-core part, 1.70 GHz on the 10-core. The memory bus holds at 128-bit, taking up to 128GB of LPDDR5x at 152GB/s.
The NPU is the headline: the identical 80-TOPS Hexagon unit from the flagship X2 Elite, nearly double the prior generation’s 45 TOPS, with FP8 in tow. Qualcomm calls it the highest-rated NPU in a PC chip. CNET attaches two caveats — Apple’s M5 is estimated at 133 TOPS, and Qualcomm declines to report “platform TOPS” the way Intel and AMD do, so rivals may still come out ahead on particular AI operations.
How to Read the Performance Numbers
The headline figures from Qualcomm and the hands-on coverage:
- Against the prior Snapdragon X Plus: single-core up to 35% faster, multi-core 17% (Qualcomm’s own numbers)
- At equal power versus Intel’s Core Ultra 7 265U: up to 3.5x the CPU performance, with Intel burning 4x to 4.6x the power
- Against the Core Ultra 7 256V: 52% faster multi-core
- UL Procyon AI score of 4,193 — better than double the 256V and more than six times the 265U; Geekbench AI at 83,624 versus 48,041 and 13,615
- Power draw up to 43% below the previous generation, with Qualcomm advertising “multiday” battery life
CNET ran Geekbench 6.5 on pre-production hardware at the venue and reproduced Qualcomm’s published figures, beating the Core Ultra 7 265U and 256V at matched power; Windows Central’s review logged no throttling and no fan noise under sustained load. Pre-production benchmarks always deserve a skeptical margin, but the direction holds together: for the first time, the Plus tier’s ceiling covers what mainstream x86 machines actually do all day.
The Mainstream Price Band Is the Real Battlefield
Windows Central’s argument pulls no punches: the $799-to-$1,299 range is where the volume lives, and most consumers and corporate fleets never glance at $1,500-to-$2,000 flagships. With the X2 Plus closing out the ladder — Elite Extreme, Elite, Plus — OEMs can keep the full NPU specification on cheaper machines, underwriting Windows 11’s on-device AI features: Cocreator, Auto Super Resolution, Studio Effects, and a swelling share of agentic AI workloads. The low power budget also makes fanless and extra-thin designs practical while partially cushioning the memory-price inflation that AI data-center demand kicked off. The flip side: should DRAM prices keep climbing, first-wave pricing may still land above expectations.
The CES chip war actually ran on two fronts. A day earlier, AMD had devoted its stage to datacenter-class hardware — the Helios rack, the MI455X accelerators (our earlier coverage) — while Qualcomm placed its chips on the device edge. The two routes complement rather than collide: frontier models stay in the cloud, everyday inference moves onto the device.
The Enterprise Wedge: Snapdragon Guardian
The X2 Plus also ships with Snapdragon Guardian, a security-and-management layer supplying out-of-band updates, remote lock-and-wipe, and device tracking — a shot aimed squarely at Intel vPro. For enterprise IT, Arm’s tallest obstacle was never raw performance; it was manageability and compatibility. Guardian fills the management hole, leaving compatibility as the remainder: some legacy enterprise tools and GPU-heavy workflows still behave erratically under emulation, and gaming anti-cheat support retains gaps (Fortnite runs now, but the list is not complete). Expect the race to keep moving — as Intel and AMD ship their next platforms, several of these gaps will close from the other direction.
What It Means for Developers and Product Teams
Three practical effects. First, an 80-TOPS-class NPU is about to graduate from flagship perk to standard equipment on mainstream laptops; raise the target hardware baseline for on-device inference — local AI features with low latency and controllable privacy just gained a far larger installed base to aim at. Second, if your product leans on CPU-intensive or x86-specific workloads, this is the moment to test Windows on Arm compatibility in earnest, before your users find the emulation edge cases first. Third, for procurement and product-support decisions, the performance-per-watt gap between x86 and Arm is hardening into a quantifiable cost line; in a 2026 of rising memory prices, that gap will surface directly in the spec sheets of same-price machines.